[RFC PATCH] hwmon: (peci/cputemp) Number cores as seen by host system

[Guenter Roeck]

On Wed, Feb 08, 2023 at 05:16:32PM -0800, Zev Weiss wrote:
> While porting OpenBMC to a new platform with a Xeon Gold 6314U CPU
> (Ice Lake, 32 cores), I discovered that the core numbering used by the
> PECI interface appears to correspond to the cores that are present in
> the physical silicon, rather than those that are actually enabled and
> usable by the host OS (i.e. it includes cores that the chip was
> manufactured with but later had fused off).
> 
> Thus far the cputemp driver has transparently exposed that numbering
> to userspace in its 'tempX_label' sysfs files, making the core numbers
> it reported not align with the core numbering used by the host system,
> which seems like an unfortunate source of confusion.
> 
> We can instead use a separate counter to label the cores in a
> contiguous fashion (0 through numcores-1) so that the core numbering
> reported by the PECI cputemp driver matches the numbering seen by the
> host.
> 

I don't really have an opinion if this change is desirable or not.
I suspect one could argue either way. I'l definitely want to see
feedback from others. Any comments or thoughts, anyone ?

> Signed-off-by: Zev Weiss <zev at bewilderbeest.net>
> ---
> 
> Offhand I can't think of any other examples of side effects of that
> manufacturing detail (fused-off cores) leaking out in
> externally-visible ways, so I'd think it's probably not something we
> really want to propagate further.
> 
> I've verified that at least on the system I'm working on the numbering
> provided by this patch aligns with the host's CPU numbering (loaded
> each core individually one by one and saw a corresponding temperature
> increase visible via PECI), but I'm not sure if that relationship is
> guaranteed to hold on all parts -- Iwona, do you know if that's
> something we can rely on?
> 
> This patch also leaves the driver's internal core tracking with the
> "physical" numbering the PECI interface uses, and hence it's still
> sort of visible to userspace in the form of the hwmon channel numbers
> used in the names of the sysfs attribute files.  If desired we could
> also change that to keep the tempX_* file numbers contiguous as well,
> though it would necessitate a bit of additional remapping in the
> driver to translate between the two.

I don't really see the point or benefit of doing that.

Thanks,
Guenter

> 
>  drivers/hwmon/peci/cputemp.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/hwmon/peci/cputemp.c b/drivers/hwmon/peci/cputemp.c
> index 30850a479f61..6b4010cbbfdf 100644
> --- a/drivers/hwmon/peci/cputemp.c
> +++ b/drivers/hwmon/peci/cputemp.c
> @@ -400,14 +400,15 @@ static int init_core_mask(struct peci_cputemp *priv)
>  static int create_temp_label(struct peci_cputemp *priv)
>  {
>  	unsigned long core_max = find_last_bit(priv->core_mask, CORE_NUMS_MAX);
> -	int i;
> +	int i, corenum = 0;
>  
>  	priv->coretemp_label = devm_kzalloc(priv->dev, (core_max + 1) * sizeof(char *), GFP_KERNEL);
>  	if (!priv->coretemp_label)
>  		return -ENOMEM;
>  
>  	for_each_set_bit(i, priv->core_mask, CORE_NUMS_MAX) {
> -		priv->coretemp_label[i] = devm_kasprintf(priv->dev, GFP_KERNEL, "Core %d", i);
> +		priv->coretemp_label[i] = devm_kasprintf(priv->dev, GFP_KERNEL,
> +							 "Core %d", corenum++);
>  		if (!priv->coretemp_label[i])
>  			return -ENOMEM;
>  	}
> -- 
> 2.39.1.236.ga8a28b9eace8
>